2025/07/17 by Rashtian, Navid, Yari, Alireza, Ardalani, Nahid +2
#Energy efficiency #Fault tolerance #Fog computing #Monte Carlo Tree Search #Resource allocation #Scalability #Security #Smart grid
paper · doi:10.57647/j.spre.2025.0902.10
The evolution of smart grids, driven by renewable energy integration, advanced metering infrastructure, and the proliferation of the Internet of Things (IoT), requires low-latency, scalable computing paradigms. Traditional cloud computing struggles with latency (200–500ms) and bandwidth limitations, making fog computing a promising alternative for real-time applications like demand response 200ms and distributed energy resource management 500ms. This study proposes a semi-decentralized fog computing framework that leverages the Monte Carlo Tree Search algorithm for optimized task scheduling across heterogeneous fog nodes (3000–9000 MIPS, 512 MB–2 GB memory). The framework integrates lightweight security (AES-128, OAuth 2.0, differential privacy with ε = 0.1), robust fault tolerance (97% task completion under 20% node failure), and Software-Defined Networking for dynamic orchestration. Evaluated using iFogSim and scenarios inspired by real-world environments e.g., California microgrids, the framework achieves a 5% higher task acceptance rate (AR: 0.80–0.86 vs. 0.32 for SPA), doubles delay utility (DU: 0.71–0.73 vs. 0.26), and improves the objective function (OBJ) by 50% (0.72–0.86 vs. 0.26–0.32) compared to benchmarks such as the Service Provider Algorithm (SPA) and Reinforcement Learning (RL). Energy consumption is reduced by 12% (0.48 watts/task), and deployment costs drop by 20%, 450/node, ensuring scalability and resilience in dynamic smart grid environments.